BART Station Predictions Based Upon Current Weather

For a full description of how this system works, check out my github.

Below is a daily live prediction of the number of people that will exit a given bart station. This data utilizes one year of hourly exits per station access at BART data. Alongside these ridership numbers, one year of historical weather for San Francisco was used accessed at weather data.

This historic data was fed into a MLlib gradient boosted tree model (300 trees) to predict ridership per station.

This model is updated daily. The total capacity is taken from the total number of trains that arrived at each station times 150 people per train. Alongside this, the weather pulled from midnight is used as the weather prediction.

Today, it is predicted that 271,416.421417 people will ride BART. In addition, today BART has capacity for 3,480,900 people.

Below the predictions, you can see an architecture of the system.